<p>Epistemic stances of learners impact how they use GenAI in learning science. Yet, even though these epistemic stances are so important, little research has examined how learners’ epistemic stances were linked to their use of GenAI, and how these stances changed after using GenAI in learning science. Incorporating multiple case studies, we purposefully engaged two pairs of ninth graders as they used different modes (visual, written, verbal, and gestural modes) to represent their epistemic stances in the context of socio-scientific argumentation. We also examined how their use of GenAI in socio-scientific argumentation might shape their multimodal representations of epistemic stances. One pair of students distanced themselves from GenAI bots. They changed from positioning that GenAI offers different perspectives to believing that the primary value of GenAI lay in communicating correct knowledge. In socio-scientific argumentation, this pair of students asked the GenAI chatbot to emulate a professor delivering long scientific texts. Another pair of students, as they developed competence in evaluating socio-scientific texts provided by GenAI, changed their epistemic stance from believing that GenAI’s benefit lies in offering multiple perspectives to a stance of cooperating with GenAI and evaluating its output. These findings reveal a need to use a multimodal approach to capture how students collaboratively communicate their epistemic stances when they use GenAI in learning science. By understanding their epistemic stances, teachers can provide in-the-moment responses that shape a shared epistemic agency between students and GenAI.</p>